[fork] Bump base tag to v0.26.0 and refresh the patch series - #6
Merged
Merged
Conversation
vLLM cut v0.26.0 on 2026-07-20. Both fork patches are still required: their upstream PRs merged *after* the release cut (vllm-project#47953 as b2b8f67 on 07-21, vllm-project#44993 as 0416dab on 07-24), so neither fix is in v0.26.0. Since both PRs are now merged, the patches are regenerated from the merged upstream commits rather than the PR heads — the source diffs are byte-identical to what upstream shipped (previously they differed only in line wrapping). Both apply to v0.26.0 with no fuzz; the code they touch (_maybe_share_embeddings, should_advance and its scheduler call site) is unchanged since v0.25.1. Verified: patch -p1 --dry-run (the exact command the image runs) applies both cleanly to a pristine v0.26.0 tree; the patched files compile; the patched should_advance and _maybe_share_embeddings are byte-identical to upstream/main; vllm/vllm-openai:v0.26.0 exists on Docker Hub. Not re-run on GPU hardware. Applied-on: v0.26.0 Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Lockstep bump for the upstream
v0.26.0release (cut 2026-07-20, published 2026-07-25).Are the patches still relevant? Yes — both.
Both backported PRs have now merged upstream, but after the
v0.26.0cut, so neither fix is in the release:v0.26.0?b2b8f679d, 2026-07-21merge-base --is-ancestor→ false)0416dab27, 2026-07-24The code each patch touches (
_maybe_share_embeddings,should_advanceand its scheduler call site) is byte-identical betweenv0.25.1andv0.26.0, so both bugs carry over unchanged. Nothing to drop this round — expect to drop both at the next release.What changed
DEFAULT_BASE_TAG→v0.26.0(workflow,Dockerfile.audioARG,fork/README.mdbuild example).v0.26.0— and, now that both PRs are merged, regenerated from the merged upstream commits rather than the PR heads. The source diffs are now byte-identical to what upstream shipped (they previously differed only in line wrapping).merge-base --is-ancestordrop criterion; dated static-verification entries added.FORK.md: refreshed tag references and folded the "check whether the PR already landed" step into the lockstep recipe.Integrated tree for review:
fork/v0.26.0—v0.26.0plus exactly these two patches as discrete commits.Verification
fork/scripts/refresh-patches.sh v0.26.0— both patches apply cleanly.patch -p1 --dry-run(the exact commandapply-patches.shruns in the image) against a pristinev0.26.0tree: both apply with no fuzz or offset.should_advanceand_maybe_share_embeddingsare byte-identical toupstream/main.vllm/vllm-openai:v0.26.0exists on Docker Hub (HTTP 200) — the workflow's base-image preflight will pass.AI assistance was used for this change (Claude Code).